Lecture 14 Let s put together a Manual Processor Hardware Lecture 14 Slide 1
The processor Inside every computer there is at least one processor which can take an instruction, some operands and produce a result. Processors can be operated in different ways for example as: A central processor unit (CPU) A peripheral of another computer Array Processor Graphics Processor (GPU) A manually programmed processor Stand alone calculator We will now design a manual (or externally programmed) processor. Hardware Lecture 14 Slide 2
The Block Diagram of a Processor Both the data and the instructions will be binary numbers. The processing will be a sequence of one or more steps controlled by a clock. The result will be a binary number. Hardware Lecture 14 Slide 3
An 8 bit processor architecture We will base our design on the von Neumann architecture (1945) which subdivided the components of a processor into arithmetic units and registers, and had a common input stream for data and instructions. We will carry out processing on 8 bit bytes (similar to the processors of the 1970 s). Hardware Lecture 14 Slide 4
The Actions of a Simple Processor As an example we will find the average of two numbers. Result = (A+B)/2 Because of our choice of architecture all numbers must be represented in 8 bits. Thus the sum A+B must be less than 256. We will see later on how to extend the processing to cope with larger numbers. Hardware Lecture 14 Slide 5
The Actions of a Simple Processor The following steps are carried out: 1. The first number is set up on the input lines and stored in a register (A) 2. The second number is set up on the input lines and stored in a register (B) 3. The arithmetic circuits are set up to add register A to register B. 4. The resulting sum of A and B is transferred back into register A. 5. The shifting circuits are set up to shift the contents of register A one bit to the right. 6. The result is loaded onto an output register (Res). The processor is a sequential digital circuit Hardware Lecture 14 Slide 6
Designing a Processor From the example we see that we need a number of different components to make our processor: Registers: Registers to store the input data (A) & (B) A register to store the result (Res) A one bit register to store the carry (C)(if any) A register to store the instruction (IR) Arithmetic Circuits: An 8 bit adder An 8 bit shifter Hardware Lecture 14 Slide 7
The data path diagram The example also suggests that the registers and arithmetic units must be connected in a specific way. A simple data path diagram might be: Hardware Lecture 14 Slide 8
The data path diagram Note the following about the data path diagram: There is no information about when the data transfers occur. The diagram shows only the possible paths where data can be transferred. The arithmetic and shift operations are done by combinational circuits without more registers. The function of the arithmetic circuits are controlled by the bits in the instruction register. The processor is unable to execute further operations on the results register. Hardware Lecture 14 Slide 9
The Arithmetic-Logic Unit (ALU) We now design one important component of the central processor unit - the ALU - which carries out arithmetic or logic operations on its two inputs A and B. We will design a 4-bit unit that can be used as a building block to construct ALUs of any precision. The select lines (S2, S1, S0) determine the function of A and B that appears on the output. Hardware Lecture 14 Slide 10
The ALU functions The selection bits determine which out of 8 possible functions is used. Selection Function 000 0 001 B-A 010 A-B 011 A plus B 100 A XOR B 101 A OR B 110 A AND B 111-1 When Cin = 1 three operations change: 011: Res = A plus B plus 1 010: Res = A - B - 1 001: Res = B - A - 1 Hardware Lecture 14 Slide 11
Designing the ALU An arithmetic logic unit is a simple combinatorial circuit that can be built from components that we have already designed. To provide the arithmetic functions we use adders and subtractors. Hardware Lecture 14 Slide 12
Designing the ALU The logic functions are provided by simple gates. We need a multiplexer for each bit of the ALU to make the function selection. Hardware Lecture 14 Slide 13
Designing the ALU To finish the job we need one more multiplexer to provide the carry out. Hardware Lecture 14 Slide 14
Extending the ALU to 8 bits We can now follow the functional approach and extend our ALU to 8 bits Similarly we can scale it up to 32 bits or larger as the need arises. Hardware Lecture 14 Slide 15
Organising the Carry We can use a 2-to-1 multiplexer to allow two different carry input bits. The Carry in may be set to 1 or to the previous Carry out. The carry may be set to zero by setting the register C to zero. Thus the full range of arithmetic operations can be used. Hardware Lecture 14 Slide 16
Problem Why set the carry this way? Wouldn t it be simpler to connect the multiplexer inputs to 1 & 0 rather than using register C? Hardware Lecture 14 Slide 17
The Shifter Earlier in the course we discussed a specialised register, called the shift register, which could shift its contents let or right depending on a multiplexer selection. We could incorporate such a shift register in our design, but instead we will make use of a simpler shifter which does not store the result. We can achieve all the functionality we need by using multiplexers, and avoid the need for complex clocking arrangements. Hardware Lecture 14 Slide 18
A four function Shifter Our first shifter design will have four functions determined by two selection bits and will have a data length of eight bits. Hardware Lecture 14 Slide 19
The Shifter Functions The basic shifter will perform four operations depending on its two control inputs. These are: 00 Hold 01 Arithmetic shift left (with carry in) 10 Arithmetic shift right 11 Rotate right It is implemented simply using one multiplexer per bit. Hardware Lecture 14 Slide 20
Arithmetic Shifts Arithmetic Left Shift Arithmetic Right Shift Rotate Right Hardware Lecture 14 Slide 21
The four function shifter circuit 00 (A) 01 (B) 10 (C) 11 (D) Hold Arithmetic Shift Left Arithmetic Shift Right Rotate Right Hardware Lecture 14 Slide 22
Adding more functions to the Shifter There are many other different possible shifts that programmers may want to use. One example is: Logical Shift Right: This could be done by using first an ALU operation (Data AND 11111110) followed by a rotate right, but that would involve more processing steps and therefore be slow. Hardware Lecture 14 Slide 23
The eight function shifter A variety of useful shifts are defined in this table. F[2] F[1] F[0] Shift Carry Function 0 0 0 unchanged 0 0 1 left rotate left 0 1 0 left 0 arithmetic left shift 0 1 1 left Cin left shift with carry 1 0 0 right rotate right 1 0 1 right 0 logical right shift 1 1 0 right Input[7] arithmetic right shift 1 1 1 right Cin shift right with carry Hardware Lecture 14 Slide 24
The circuit of the eight function shifter It is a trivial matter to design a similar shifter of any precision. Hardware Lecture 14 Slide 25
The Data Path Diagram Again We now can put more detail onto the data path diagram: Note that we have not used the Cin input to the shifter in this processor. It will be connected to logic-0. Hardware Lecture 14 Slide 26